Scale the ProteomeXchange full-text pilot to open-access proteomics papers from 2020 to mid-2025 #1

What to do

Run the method of pilot/pilot_fulltext.py on all open-access papers in Europe PMC from 2020 to mid-2025 that mention ProteomeXchange. Take accessions from the full text, check each with the PROXI API, and report counts by year and by archive.

Before scaling up, add what the pilot lacks: a contact email in the User-Agent, a cache of downloaded pages and API answers, and small batches with pauses. Keep per-dataset results out of the repository.

Done when

• A script in pilot/ (or a new src/) runs the full set.
• results/<date>.md gives, for each year and each archive: papers read, accessions found, datasets still unreleased, and papers that cite one.
• No file naming papers or accessions is committed.

How the result is checked

• Anyone can re-run the script and get the same counts, apart from papers Europe PMC has added since.
• The check workflow passes, so no accession number is in the repository.
• A second agent or person re-queries a random sample of 20 unreleased datasets with PROXI and confirms their status.

Context

The pilot read 500 papers from January to March 2025 and found 10 of 694 datasets unreleased (results/pilot-2026-10-19.md). Europe PMC's annotations miss unreleased datasets, so the full text is needed. The policy in README.md allows counts only.

Automated posting by agentic team with human oversight.

Implementation pull request: PR #9.